3694
Biochemistry 1993, 32, 3694-3707
Structural Characterization of the Disulfide Folding Intermediates of Bovine a-Lactalbumin? Jonathan J. Ewbankt and Thomas E. Creighton' European Molecular Biology Laboratory, Posrfach 10.2209, Meyerhofstrasse 1, 6900 Heidelberg, Germany Received October 26, 1992; Revised Manuscript Received January 25, 1993
ABSTRACT: Specific three- and two-disulfide intermediates that accumulate transiently during reduction of the disulfide bonds of Ca2+-bound bovine a-lactalbumin have been trapped, isolated, and characterized. The three-disulfide intermediate was shown to lack the Cys6-120 disulfide bond, confirming the observations of others. The newly-recognized two-disulfide form has been shown to lack the Cys6-120 and Cys28-111 native disulfide bonds. The remaining native disulfide bonds in the two partially reduced derivatives of a-lactalbumin are stable only when the proteins are in a Ca2+-bound state. Otherwise, they adopt an equilibrium between molten globule and unfolded conformations, and rapid thiol4isulfide interchange occurs, at a rate as high as when the proteins are fully unfolded in 8 M urea, to generate distinct mixtures of rearranged products. Urea gradient electrophoresis, circular dichroism, fluorescence, and ANS binding have been combined to give a detailed structural picture of a-lactalbumin, its derivatives with native and with nonnative disulfide bonds, and the fully reduced protein. The native structure of a-lactalbumin appears to be split by selective disulfide bond cleavage into at least one subdomain, which retains the Ca2+-binding site. The a-lactalbumin molten globule state is shown largely to result from nonspecific hydrophobic collapse, to be devoid of cooperative or specific tertiary interactions, and not to be stabilized substantially by the native or rearranged disulfide bonds.
Single-domain proteins usually unfold reversibly by a cooperative process in which partially-folded intermediates between the fully folded, native state and the fully unfolded state are not stable at equilibrium. The small, but growing, number of proteins that are found to be exceptions to this rule appear under particular conditions to have conformational properties similar in some respects to those of the native state and otherscharacteristicof the unfolded state. This has usually been interpreted in terms of a single conformational state, the so-called molten globule (MG),' intermediate between the native and unfolded states but otherwise poorly understood [see Kuwajima (1989), Baldwin (1991), and Ptitsyn (1992) for reviews]. This conformational state is especially important because a wide variety of globular proteins appear to pass through a transient MG-like state during folding (Ptitsyn et al., 1990). Detailed characterizationof a stable molten globule J.J.E. acknowledges receipt of an EMBL predoctoral fellowship. Present address: Department of Biology, McGill University, 1205 Dr. Penfield Avenue, Montreal, PQ, Canada H3A 1B1. I Abbreviations: aLA, a-lactalbumin with its four native disulfide bonds intact; apo-aLA, aLA without liganded Ca2+; 3SS, bovine a-lactalbumin with the Cys6-120 disulfide bond reduced; 2SS, bovine a-lactalbumin with Cys6-120 and Cys28-111 disulfide bonds reduced; R, fully reduced bovine a-lactalbumin;[nSS] mixture of disulfide isomers other than nSS containing n disulfide bonds, derived from nSS by intramoleculardisulfide rearrangements;[nSSIu,mixture of isomers other than nSS containing n disulfide bonds, derived from nSS in 8 M urea by intramolecular disulfide rearrangements; ANS, 1-anilinonaphthalene8-sulfonate; BPTI, bovine pancreatic trypsin inhibitor; CD, circular dichroism; DTT;;, reduced dithiothreitol; DTT;, oxidized dithiothreitol; EDTA, ethylenediaminetetraaceticacid; HPLC, high-performance liquid chromatography;I-AEDANS, N-iodoacetyl-N'-(5-sulfo-1-naphwavelength of maximum fluorescence emisthy1)ethylenediamine; A,, sion; MG, molten globule; N , native, fully folded conformation; NMR, nuclear magnetic resonance;TFA, trifluoroacetic acid; U, fully unfolded conformation;UV, ultraviolet. Intermediates with all thiol groupstrapped with iodoacetamide, iodoacetate, or I-AEDANS are designated by subscripts cam, cm, or AEDANS, respectively. For partially blocked intermediates, the blocked thiol groups are indicated, for example: 2SS6/12kam, 2SS with Cys6 and Cysl20 carboxamidomethylated.
0006-2960/93/0432-3694%04.00/0
state should therefore be relevant to the more general problem of how proteins fold. One of the best-characterized MG states is that of bovine a-lactalbumin (Kuwajima, 1989). The native state of this protein is stabilized by binding of a single Ca2+ion (Hiraoka et al., 1980; Acharya et al., 1989, 1991). When the Caz+ is dissociated at acid pH, at elevated temperatures, or at intermediateconcentrationsof denaturant, aLA tends to adopt the MG conformation; it is compact (Dolgikh et al., 1981) and, as judged by NMR (Baum et al., 1989) and far-UV CD [e.g., Kronman et al. (1965); Kuwajima et al. (1981)], it contains some of the secondary structure present in the native protein. But the native tertiary interactions appear to be greatly diminished: the near-UV CD spectrum is greatly reduced, and there are few, if any, native tertiary interactions detectable by NMR (Baum et al., 1989). The a-lactalbumin MG state present at equilibrium appears to be similar to a kinetic intermediate state that is populated transiently during refolding (Kuwajima et al., 1985; Ikeguchi et al., 1986a). Partially folded intermediatesthat are intrinsically unstable can be trapped in a stable form if their folding is coupled to disulfide bond formation. Native a-lactalbumin has four disulfidebonds, Cys6-120, Cys28-111, Cys61-77, andCys7391, when it is designated here as aLA. The results of an extensive kinetic analysis of disulfide bond formation and reduction in the bovine protein are reported in the preceding paper (Ewbank t Creighton, 1993). When aLA is folded, reductive cleavage of one of its four disulfide bonds, that between Cys6 and Cysl20, occurs at a rate hundreds of times faster than that of any of the other three disulfides, and a unique three-disulfide intermediate (3SS) is produced (Shechter et al., 1973; Iyer & Klee, 1973). Ca2+-bound3SS maintains a folded conformation (Kuwajima et al., 1990) and is reduced specificallyto generate a two-disulfide intermediate (2SS) that accumulates if Ca2+ is present at high concentrations (Ewbank & Creighton, 1993). This intermediate, which had not previously been observed, has been trapped 0 1993 American Chemical Society
a-Lactalbumin’s Disulfide Folding Intermediates and isolated, and the second cleaved disulfide bond is identified here as being that between Cys28 and Cysl 11. The native disulfidesof the homologousthree-disulfide form of human a-lactalbumin have been shown previously to be stable when the protein is in its native conformation, with Ca2+bound (Ewbank & Creighton, 1991). Otherwise, when human 3SS adopts a molten globuleor unfolded conformation, its disulfides rearrange intramolecularly. The apo forms of both 3SS and 2SS of the bovine protein are also shown here to adopt molten globule and unfolded types of conformations; their remaining native disulfides are then unstable and spontaneously rearrange intramolecularly. The structures of bovine aLA, the two intermediates 3SS and 2SS, together with their rearranged derivatives and the fully reduced protein (R), have been investigated directly by CD and intrinsic fluorescence, by transverse urea gradient gel electrophoresis,and by binding of the hydrophobic fluorescent dye ANS. The characterization of these forms has allowed a fuller understanding of the pathways of reduction and reformation of a-lactalbumin’s disulfide bonds and of the stabilization of its different conformational states. In particular, it is shown that the native disulfides do not stabilize the MG state substantially, and vice versa, consistent with the tendency of the native disulfides of 3SS to rearrange when it is in this conformation. The partly folded conformation of Caz+-bound 2SS further illustrates the ability of selective disulfide bond cleavage to dissect folded proteins into stable subdomains. MATERIALS AND METHODS
Materials. Trypsin (type XIII, TPCK-treated) and IAEDANS (N-iodoacetyl-N’-(5-sulfo-1-naphthyl)ethylenediamine) were obtained from Sigma, and ANS was from Fluka. They were used without further purification. All other materials were as described elsewhere (Ewbank & Creighton, 1993). Preparation of Trapped Fully Reduced and Partially Reduced Forms of a-Lactalbumin: ( a ) Blocked Fully Reduced a-Lactalbumin (Ream and Rem). DTT;: (100 mM) in 0.1 M Tris (pH 8.7), 0.2 M KCl, 1 mM EDTA was added to an equal volume of aLA (40 mg/mL in the same buffer). After 15 min of incubation at 25 OC, when the protein was fully reduced, all free thiols were blocked by the addition of l / 5 volume of 0.6 M iodoacetamide or of 0.6 M iodoacetate (neutralized with KOH), in 0.5 M Tris (pH 8.0). After 2 min of further incubation at 25 OC, the protein was separated from the excess blocking reagent by rapid gel filtration into 20 mM ammonium acetate (Pharmacia Fast Desalting column) and then lyophilized. ( b )Iodoacetamide- Trapped Two-Disulfide a-Lactalbumin (2SScam). 2SS was purified by reversed-phase HPLC as described in the precedingpaper (Ewbank & Creighton, 1993). After the addition of 1 M CaCl2 to 20 mM, the free thiols were trapped with 0.6 M iodoacetamide as above. The protein was rechromatographed by reversed-phaseHPLC (Dynamax300A C4 column 4.6 X 250 mm, with a linear gradient of 25% to 50% acetonitrile in 0.1% TFA) and lyophilized. ( c ) Iodoacetamide- Trapped Three-Disulfide a-Lactalbumin (3SSCam).One-tenth volume of 60 mM DTT;: in 0.1 M Tris (pH 7), 0.2 M KC1, 10 mM CaC12 was added to a 20 mg/mL solution of aLA in the same buffer. After 2 min of incubation at 25 OC, all free thiols were trapped with 0.6 M iodoacetamide as detailed above. The mixture was desalted, and then 3SS,, was further purified by reversed-phaseHPLC as above and lyophilized.
Biochemistry, Vol. 32, No. 14, 1993 3695 Preparation of Trapped Rearranged-Disulfide Forms of a-Lactalbumin. ( a ) Iodoacetamide- Trapped Rearranged Two-Disulfide a-Lactalbumin ([ZSSIcam and [ZSS];,,). Lyophilized acid-trapped 2SS with free thiols, prepared as describedin the preceding paper (Ewbank BE Creighton, 1993), was dissolved to a concentration of approximately 120 pM in 10 mM HCl and then diluted 24-fold into either 0.1 M Tris (pH 8.7), 0.2 M KC1, 1 mM EDTA, or the same buffer containing 8.35 M urea. After 2 min at 25 OC, the free thiols were trapped with 0.6 M iodoacetamide, as above. One-fifth volume of acetonitrile (containing 0.1% TFA) was then added, and the mixture was applied to a C4 reversed-phase column as above. The column was washed with 25% acetonitrile, 0.1 % TFA to elute the excess blocking reagent, and then all the protein was eluted by a single step to 80% acetonitrile, 0.1% TFA. The protein-containing fraction was lyophilized. The mixtures obtained in the presence and absence of urea were designated [2SS]:a, and [2SS]cam, respectively. ( b ) Iodoacetamide- Trapped Rearranged Three-Disulfide a-Lactalbumin ([3SS],,, and [3SS]:a,). 3SS was generated by partial reduction of Ca2+-boundaLA and then diluted into an EDTA- or urea-containing buffer as described in the preceding paper (Ewbank & Creighton, 1993). After 2 min at 25 OC, the protein was trapped with 0.6 M iodoacetamide and purified in the same way as [2SS]ca, as above. The mixtures obtained in the presence and absence of urea were designated [ 3SS]:a, and [3SS],am, respectively. Preparation of AEDANS- Trapped Partially Reduced Forms of a-Lactalbumin. One-fourth volume of 40 mM DTT;: in 0.1 M Tris (pH 8.7), 0.2 M KC1, 10 mM CaC12 was added to aLA at 12 mg/mL in the same buffer. After 2 min of incubation of 25 OC, the mixture of 3SS, 2SS, and R was trapped by the addition of l / 5 volume of 0.2 M I-AEDANS in 0.5 M Tris (pH 8.0), to modify all free thiol groups covalently. After 10 min at room temperature, the AEDANS-trapped proteins were separated by reversed-phase HPLC as above, and the protein-containing fractions were lyophilized. Peptide Mapping. The AEDANS-trapped proteins were resuspended in 0.2 M ammonium acetate, 100 pM CaC12, pH 7.8, at a concentration of approximately 0.5 mg/mL. Onethird volume of a 1 mg/mL solution of trypsin in 0.2 M ammonium acetate, 10 mM EDTA, pH 7.8, was added, and the mixture was incubated at 25 OC for 2.5 h. The resulting peptides were separated by reversed-phase HPLC (Dynamax300A C4 column 4.6 X 250 mm, with a linear gradient from 0% to 56% acetonitrile in 0.1% TFA) and then subjected to amino acid analysis. CD Measurements. aLA and its covalently-trapped disulfide derivatives were dissolved to a concentration of between 5 and 25 pM in 0.1 M Tris (pH 8.7), 0.2 M KCl that also contained either 10 mM CaC12 or 1 mM EDTA (neutralized with KOH). For spectra of the unfolded proteins, the buffer also contained 8 M urea. CD spectra were acquired at 25 OC on a Jobin Yvon CD6 spectrometer calibrated with (+)-locamphorsulfonic acid, using cells of path length 10 mm and 1mm, with subtraction of the buffer contribution. The results were expressed as mean residue ellipticity (deg*cm2-dmol-l), calculated using mean residue weights for aLA, 3SS,m, 2SSCa,, and % , of 115.4, 116.2, 117.3, and 119.2, respectively. Protein concentrations were calculated using an extinction coefficient at 280 nm of 28 500 M-I cm-I for aLA and 27 200 M-I cm-l for &, (Ikeguchi & Sugai, 1989). The value for % ,, was taken to be the same as for &m and those for 3SSca, and 2SSca, were assumed to be 28 000 and 27 500 M-l cm-l, respectively. The values for the rearranged
3696 Biochemistry, Vol. 32, No. 14, I993 derivatives were taken to be the same as for the form with the corresponding number of native disulfide bonds. CD spectra were also recorded for unblocked and partially blocked fully reduced a-lactalbumin. One-tenth volume of 100 mM DTT;: in 0.1 M Tris (pH 8.7), 0.2 M KCl, that also contained either 10 mM CaCl2 or 1 mM EDTA (neutralized with KOH), was added to aLA at approximately 15 pM in the same buffer. After 15 min of incubation at 25 OC, when the protein was fully reduced, spectra were acquired as above. For spectra of the unfolded protein, the buffer also contained 8 M urea. Spectra of Rb/laocam and R;[ji:E:m were also acquired in an identical fashion, but for the substitution of 3SScam and 2SSam for aLA, respectively. For all measurements in solution, acquisition of data was conducted as rapidly as possible with freshly dissolved protein solutions. An absence of time-dependent changes and lack of significant absorbance above 320 nm indicated the absence of aggregation. ANS-Binding Studies. aLA and its covalently-trapped disulfide derivatives were dissolved to 25 pM in 0.1 M Tris (pH 8.7), 0.2 M KCl, 50 pM ANS, containing either a fixed concentration of CaCl2 or 1 mM EDTA. The concentration of ANS was determined using an extinction coefficient of 5000 M-' cm-' at 350 nm (Weber & Young, 1964). The protein concentration was varied by serial dilution in the same buffer, thereby keeping the ANS concentration constant and so avoiding inner filter effects (Mulqueen & Kronman, 1982). The fluorescence emission at 490 nm was measured at 25 OC using an excitation wavelength of 350 nm, with correction for the contribution of the buffer. Intrinsic Fluorescence. aLA and its covalently-trapped disulfide derivatives were dissolved to 1.8 pM in 0.1 M Tris (pH 8.7), 0.2 M KC1, containing either 10 mM CaCl2 or 1 mM EDTA. The emission spectra were recorded at 25 OC between 300 and 450 nm with bandwidths of 8 nm using excitation at 280 nm, and with correction for the contribution of the buffer. Urea Gradient Gel Electrophoresis. Transverselinear urea gradient polyacrylamide slab gels were prepared essentially as described by Goldenberg and Creighton (1984), with the use of a discontinuous high pH buffer system [gel buffer, Tris-HC1at pH 8.8; electrode buffer, Tris-glycine at pH 8.3; Davis (1964)], modified to include either 5 mM CaC12 or 1 mM EDTA in both the gel and the running buffer. The gradient of urea was 0-8 M and the compensating inverse acrylamide gradient was 14% to 11.5%. The electrode buffer was only used once (Ewbank & Creighton, 1993). Electrophoresis was at 25 OC at a constant current of 30 mA/gel.
RESULTS Reduction of Ca2+-boundaLA proceeds through specific three- and two-disulfide intermediates, 3SS and 2SS, respectively (Ewbank & Creighton, 1993). With Ca2+bound, the disulfide bonds of 3SS and 2SS are relatively resistant to reduction and stable with regard to disulfide bond rearrangements. If, however, 3SS or 2SS were fully unfolded by the addition of 8 M urea, their disulfide bonds rapidly rearranged to give complex mixtures of disulfide bond isomers, [3SSIu and [2SSIu,respectively. Removal of bound Ca2+from 3SS or 2SS in the absence of denaturant also resulted in rapid disulfide bond rearrangements that generated, however, distinct mixtures of disulfide bond isomers, denoted as [3SS] and [2SS], respectively. As these intermediates have both disulfides and free thiols, they must be trapped in a stable form to be structurally characterized. There are two commonly used trapping methods: acid quench (which greatly
Ewbank and Creighton reduces thiol reactivity by minimizing the reactive thiolate anion) and irreversible covalent modification of thiol groups, with iodoacetamide for example. Both methods were used for the kinetic analysis reported in the preceding paper and gavevery similar results (Ewbank & Creighton, 1993). Acidtrapped intermediates can only be characterized at acid pH, however, where the single Ca2+ ion dissociates from aLA (Permyakov et al., 1981) and radically changes the protein's conformation (Kuwajima, 1989;Goto et al., 1990). Therefore, intermediates were trapped by covalent modification, which has the advantage of being irreversible and allowed characterization of the intermediates under the same conditions as used for the studies on disulfide formation and breakage. The blocking groups introduced may, however, influence the conformational properties of the protein, which is discussed further below. aLA and all its derivatives were greater than 95% pure as judged byreversed-phase HPLC and native gel electrophoresis. The disulfidecounting method describedin the preceding paper (Ewbank & Creighton, 1993) was used to confirm that each species had the expected number of disulfide bonds and that no significant oxidation, reduction, or intermolecular association had occurred during the rearrangement of the twoand three-disulfide derivatives. Identifying the Disulfide Bonds Broken in 2SS and 3SS. The intermediate 3SS is well-established and known to have the Cys6-120 disulfide broken (Shechter et al., 1973; Kuwajima et al., 1990). The intermediate 2SS in which two disulfide bonds had been cleaved had not previously been observed, so it was necessary to identify the second disulfide that had been reduced. This was accomplished by isolating the intermediate and performing peptide mapping. Reduction by DTT;: of aLA in the presence of 10 mM Ca2+ was quenched, by the addition of I-AEDANS, when both 3SS and 2SS were present, and the two trapped proteins (3SSAEDANS and 2SSAEDANS) were purified to homogeneity. The two proteins were digested with trypsin to determinewhich cysteine residues were blocked by AEDANS. AEDANS has substantial absorbance at 340 nm, so those peptides that contained an AEDANS-modified cysteine residue could be readily identified. A comparison of the products of tryptic digestion Of ~SSAEDANS and ~SSAEDANS is shown in Figure 1. The identity of the different peptides was established by amino acid analysis and reference to a-lactalbumin's primary structure (Shewale et al., 1984; Vilotte et al., 1987). Two AEDANS-containing peptides were obtained from 3SSAEDANS: tryptic fragments Cys6 to ArglO and Leu1 15 to Lys122. These peptides each contained a single cysteine residue and confirmed that the disulfidelinking Cys6 to Cysl20 was cleaved in 3SS. The tryptic digest of 2SSAEDANS had the same two AEDANS-containing peptides, plus two additional AEDANS-containingpeptides identified as Gly 17-Phe3 1and Alal09-Lysll4 (Figure 1). These peptides also each contain a single cysteine residue and indicate that the second disulfide bond cleaved was Cys28-111, Urea Gradient Gel Electrophoresis. During transverse linear urea gradient electrophoresis, a protein migrates electrophoretically through a slab gel of polyacrylamide in which there is a linear gradient of urea perpendicular to the direction of migration. The rate of migration is sensitive to both the net charge on the protein and the degree of unfolding induced by the urea; compact species migrate more rapidly than their unfolded forms. The technique can therefore provide direct visualization of the dependence of protein conformation on denaturant concentration, and mixtures of species can be analyzed directly. The discontinuous buffer
a-Lactalbumin's Disulfide Folding Intermediates
Elution volum
FIGURE 1: Comparison of the tryptic peptides from 3SSAEDANS (top) and 2SSAEDANS (bottom). Peptides were eluted from reversed-phase HPLC using the absorbance at 340 nm to identify those labeled with AEDANS. Amino acid analysis identified the peptides as being CysbArglO (1) and Leu115-Lys122 (2), Ala109-Lys114 (3) and Gly17-Phe31 (4). The peptide Gly17-Phe31 presumably resulted from cleavage at an atypical peptide bond and was not recovered in a stoichiometric yield.
system (Davis, 1964) used here runs at pH 9.5, close to that used here in the kinetic studies of disulfide bond reduction and re-formation (Ewbank & Creighton, 1993) and in the other physical studies reported here. Binding of Ca2+ will affect the electrophoretic mobility because of the charge of the bound ion; there may also be effects caused by accompanying changes in the protein's conformation. Consequently, the urea-gradient gels give a wealth of information about the relative hydrodynamic volumes of the various species, both when folded and unfolded, about their binding of Ca2+,and about the stabilities of their folded conformations and the relative cooperativity of their unfolding transitions. The patterns obtained with aLA, 3SSca,, 2SScam, and &am in the presence and absence of Ca2+ are given in Figure 2. The results for aLA are consistent with previous results (Creighton, 1979), obtained before aLA's Caz+-bindingwas recognized (Hiraoka et al., 1980) and when the complex nature of its unfolding transition was not fully explicable. Fully reduced a-lactalbumin does not have a measurable Ca2+ affinity, so the mobilities of the trapped species R,, (Thompson et al., 1988) and &am are not affected by Ca2+, and gels run under the two conditions could be compared (Table I). The presence of EDTA per se should not alter the mobility of a-lactalbumin, as the proposal that EDTA binds to the protein (Kronman & Bratcher, 1983) has been convincingly repudiated (Stemmer & Klee, 1990, and references cited therein). &, had a constant mobility across the whole gradient, indicative of its being fully unfolded, even in the absence of
Biochemistry, Vol. 32, No. 14, 1993 3697 urea. By contrast, the mobility of &m decreased gradually with increasing urea concentration, indicative of gradual unfolding, which was the same in the presence and absence of Ca2+. In 8 M urea, where both &, and &am are fully unfolded, their electrophoreticmobilitiesare assumed to differ only due to the negative charge of the eight acidic cm groups of &,; in that case, their relative mobilities indicate that the net charge of the apo forms of aLA and of the other iodoacetamide-trapped forms was approximately -1 2 under theconditions of electrophoresis,which is consistent with their content of acidic and basic residues. Various degrees of unfolding were observed for aLA and its three- and two-disulfide derivatives, as judged by their decreasing mobilities at high urea concentrations, both in the presence and absence of Ca2+. In each case the curves leveled off before the end of the urea gradient, indicative of completion of the unfoldingtransition, A continuous electrophoreticband was always obtained, indicating that all the folding-unfolding transitions were rapid relative to the time of electrophoresis. In the presence of EDTA, aLA, 3SS,m, 2SS,m, and &am each had distinct electrophoretic mobilities and underwent gradual decreases in mobility with increasing urea concentration (Figure 2); it was difficult to define start and end points for these broad unfolding transitions. Trapping with iodoacetamide does not introduce any charges, so the differences in mobility should reflect primarily differences in hydrodynamic volume, with aLA being the most compact. The different mobilities of the unfolded proteins reflected the extent to which they were constrained against complete expansion by their disulfide bonds. The change in mobility appeared to be approximately proportional to the size of loop broken by each successive cleavage. Thus, cleavage of the Cys6-120 disulfide opened a 3 1-residueloop and produced a significant retardation of unfolded 3SScam relative to the unfolded aLA, and cleavage of Cys28-111 and of the last two disulfides (Cys61-77 and Cys73-91) opened 63- and 30residue loops, respectively, with proportionate consequences for the mobilities of 2SSCa, and &am+ Retention of disulfide bonds in the fully unfolded protein decreased the change in mobility upon unfolding of aLA and, to lesser extents, 3SS,,, and 2SS,m, and consequently decreased the apparent abruptness of their unfolding transitions. aLA, 3SS,,,, and 2SS,, all bind Ca2+with relatively high affinity in the absence of denaturant (Ewbank & Creighton, 1993). As expected, the electrophoreticmobilities of all three species were changed at low urea concentrations by the inclusion of Ca2+ during electrophoresis. Their mobilities were then indistinguishable, indicating that they all had very similar hydrodynamic volumes and adopted a native-like compactness. No protons are released from aLA upon Ca2+ binding (Schaer et al., 1985),so Ca2+-boundaLA, 3SS,m, and 2SS,, should each have an overall charge less negative by two units relative to Ca2+-free (apo) aLA and a decreased electrophoretic mobility (see Figure 3 for a schematic interpretation). The decrease in mobility of aLA upon binding Ca2+ was close to that expected if due only to a change in its net charge from -12 to -10. The decrease in mobility of 3SS,, was not so great as for aLA, so there was probably a compensating compaction of 3SS,,, upon binding Ca2+. In the case of 2SS,,, the protein's mobility was only slightly altered in the presenceof CaZ+(Figure 2), so compaction of the conformation on binding Ca2+nearly compensated for the increasein charge. The mobilities of aLA, 3SS", and 2SS,,, were not altered by CaZ+in the presence of 8 M urea, so the unfolding of the proteins was accompanied by dissociation of bound Ca2+,
3698 Biochemistry, Vol. 32, No. 14, 1993
Ewbank and Creighton
- Rcm - Rcam - 2sscam - 3sscam -.1
w
OM
A
8M
[Urea]
FIGURE2: Urea gradient electrophoresis of a L A and its trapped reduction products in the presence (top) and absence (bottom) of Ca2+.A 100-pL sample of a mixture containing each component (at a concentration of approximately 20 pM) was layered on the top of each gel. The upper gel contained 5 mM CaC12 and the lower one contained 1 mM EDTA. The gels were run at 25 "C at a constant 30 mA for 4 h and 2.5 h, respectively. The arrow indicates the direction of migration. For the lower gel, electrophoresis was stopped after 1.7 h, 100 pL of 20 pM Rcm was applied, and the electrophoresis continued. In this way, the band of Rcm did not overlap any of the other bands. The discontinuity at the 0 M urea end of the gradient of the lower gel was caused by a polymerization artefact. Table I: Effects of Ca2+and Urea on the Relative Electrophoretic Mobilities of aLA and Its Trapped Disulfide Derivatives during Discontinuous Polyacrylamide Gel Electrophoresis at pH 9.5 and 25 "C relative mobilitv species aLA 3SScam 2SScam Rcam
Rem'
1 mM EDTA
100 92 81 63 80
5 mM Ca2+ 80 80 80 63 80
8 M ureab 78 70 57 48 80
The charge of the Ca2+ion decreases the net negative charge of each speciesthat binds it. The mobility of each species was unaffected by the presence or absence of Ca2+when unfolded by 8 M urea. Rcm has a net negative charge 8 units greater than the apo forms of the other species. (I
consistent with previous results (Permyakov et al., 1985). Consequently, the unfolding transitions observed in the presence of Ca2+ were complicated by the twin effects of changesof conformationand of charge upon Ca2+dissociation (Figure 3). This meant that the decrease in mobility caused by unfolding was compensated in part by an increasein mobility that resulted from the dissociation of the positively charged Ca2+ligand from the unfolded protein. As a result of this effect and the constraints of the disulfide bonds on the hydrodynamic volume of the fully unfolded conformation, the unfolding transitions in the presence of Ca2+did not appear particularly abrupt, although they were much more so than those of the apoproteins. The midpoint of the unfolding transitions for each species occurred at varying urea concentrations in the order aLA > 3SScam > 2SScam.
The unfolding of aLA by denaturants is dependent upon Ca2+concentration (Kuwajima et al., 1986; Ikeguchi et al., 1986b). The maximum concentration of Ca2+that could be included during electrophoresis was 5 mM. In such Ca2+containing gels, a semicrystalline front was observed to run by the dye front presumably due to the migration of Ca2+ions during the course of electrophoresis. Consequently, the Ca2+ concentration in the gels may have varied during the course of electrophoresis, and no rigorous analysis of the unfolding transitions was attempted. Under these conditions, the mobilities of aLA, 3SScam, and 2SScam increased with increasing, but low, urea concentration, probably due to the affinity of the native state for Ca2+being reduced by urea, shifting the equilibrium between Ca2+-boundand apo forms of each. The electrophoretic patterns, under nondenaturing conditions, for the trapped rearranged two- and three-disulfide derivatives of aLA are shown in Figure 4b. In each case there appeared to be many different rearranged species that were poorly resolved by electrophoresis or by HPLC (Figure 4a), as was observed previously for the rearranged threedisulfide forms of human a-lactalbumin (Ewbank & Creighton, 1991). The trapped products of rearrangement of 3SS generated in the absence of urea ([3SS]cam) tended to have greater electrophoretic mobilities than those generated in the presence of urea ( [ 3SS]:am);similar observations were made of the products of rearrangement of 2SS. A proportion of [3SS]mmhad a mobility greater than 3SS,m, with some species surprisingly having mobilities as great as apo-aLA. The majority of the [2SS]cam species were more compact than 2SScam as judged by their electrophoretic mobilities. All the
Biochemistry, Vol. 32, No. 14, 1993 3699
a-Lactalbumin’s Disulfide Folding Intermediates a.
4
. Elution volume
FIGURE3: Schematic representation of the factors influencing a protein’s electrophoretic mobility. The hypothetical protein with charge [-z] binds CaZ+(solid circle) only when in a fully folded conformation. Removal of the ligand alters the protein’s overall charge and hence its mobility (l), but may also alter the protein’s conformation (2). Unfolding of the Ca2+-bound protein is accompanied by a change in charge due to the dissociation of the ligand, which in this case compensates in part for the simultaneous change in conformation (3). The unfolded protein is constrained from complete unfolding by its two disulfide bonds (black zig-zags). When these are reduced and blocked with neutral iodoacetamide, there is no change in charge, but the protein’s mobility is further decreased (4). Blocking thereduced thiolswithiodoacetate increases thecharge and hence the reduced protein’s mobility ( 5 ) . The large arrow indicates the direction of increasing mobility. For fully reduced a-lactalbumin, the nature of the blocking group influences the conformation of the fully reduced protein, but this added complexity has been omitted from this scheme.
species, including [2SS]L,, had mobilities substantially greater than Ram. Essentiallyidentical patterns were obtained for the trapped rearranged proteins after electrophoresis in the presence of 5 mM CaC12, indicating that at least the majority of the rearranged species did not bind Ca2+ with even low affinity (Ewbank, 1992). Each of the mixtures gave broad electrophoretic bands and gradual unfolding transitions upon urea gradient electrophoresis (Figure 5 ) . The width of the band was not markedly greater in the transition region, indicating that the majority of species in each had similar unfolding transitions. The [3SS],,m species were slightly more stable and underwent a smaller relative decrease in mobility than did [3SS]km;the overlap between the two was less in the presence of 8 M urea. The distinction between the two forms was less pronounced in the case of the rearranged two-disulfide derivatives. Thus, all the rearranged species, even those that were generated in the presence of urea, possessed a degree of compactness in its absence. Circular Dichroism. CD can give information both about a protein’s secondary and tertiary structure. The near-UV CD spectrum contains contributions from each aromatic residue and disulfide bond in an asymmetric environment.
.
,
,
.
I
I
I
&
Electrophoreticmobility
FIGURE4: Patterns obtained after chromatographic (a) and electrophoretic (b) separation of blocked rearranged derivatives of a-lactalbumin. The mixtures are, from top to bottom, [3SS],,, [3SS]L,, [2SS],,,, and [2SS];,,. (a) Cd reversed-phase HPLC with a gradient of 25%50% acetonitrile in 0.1%TFA. aLA, 3SS,,, and 2SS,,, are poorly resolved under these conditions and all elute close to the position given by the arrowhead. (b) High-pH discontinuous polyacrylamide gel electrophoresis at 25 OC in the presence of 1 mM EDTA. The arrowheads mark the positions of, from right to left, aLA, 3SS,,, 2SS,,, and &,.
The intensity and shape of the near-UV spectrum is therefore primarily a measure of the folded tertiary conformation. A protein’s far-UV CD spectrum, on the other hand, reflects mainly the contribution of nonrandom conformations of the peptide backbone, possibly with an additional contribution from restrained aromatic residues and disulfide bonds (Manning & Woody, 1989; Bolotina, 1987). The MG conformation is characterized by a near-native content of secondary structure and little or no tertiary structure [see Kuwajima (1989)]. Consequently, the near- and farUV CD spectra can be interpreted semiquantitativelyin terms of the relative populations ofthe N, MG, and U states, although detailed interpretation of the far-UV spectra in terms of secondary structure were not attempted (Kuwajima et al., 1985;Ikeguchi & Sugai, 1989). The CDspectra wereobtained under the same conditions used for the kinetic studies of disulfide bond formation and breakage, although necessarily of disulfide intermediates that had been covalently trapped. The far-UV spectra of aLA in the native conformation, with Ca2+bound, and in the unfolded state in 8 M urea (Figure 6)closely resembled those previously reported [for example, Kuwajima et al. (1981), Gil’manshin et al. (1982),2 and Ikeguchi and Sugai (1 989)]. The near-UV spectrum of native aLA was very complex, reflecting the overlap of dichroic contributions from aLA’s 12 aromatic residues and four disulfide bonds, which are spread throughout the protein molecule. The pronounced negative ellipticity at around 270 nm and the positive peak at 295 nm have been ascribed to Tyr and Trp residues, respectively (Cowburn et al., 1972). The far-UV spectrum of native aLA was typical of a largely a-helical protein with minima at around 208 nm and 222 nm. In 8 M urea, the spectrum was consistent with the protein ~
~~
* The figure legends for Figures 1 and 2 in Gil’manshin et al. (1982) appear inverted.
3700 Biochemistry, Vol. 32, No. 14, 1993
Ewbank and Creighton
FIGURE5: Urea gradient electrophoresis in the presence of 1 mM EDTA of iodoacetamide-trapped rearranged three- and two-disulfide derivatives of a-lactalbumin. The conditions were as given in the legend to Figure 2: (a) [3SS],a,, (b) [3SS]:!,, (c) [2SS],a,, (d) [2SS]:a,. The upper band in each is R,,, applied two-thirds of the way through electrophoresis.
being fully unfolded (Kuwajima et al., 1991, and references cited therein). The near-UV spectrum was essentially featureless, and the intensity of the far-UV spectrum was greatly diminished. Ca2+-bound3SScam appeared to be vitrually fully folded as judged by CD. Its far-UV spectrum was very similar to that of aLA. The difference between the two in the near-UV was broad and featureless and could be attributed to the dichroic contribution of the broken disulfide bond (Kuwajima et al., 1990; Ikeguchi et al., 1992). Ca2+-bound 2SScam, on the otherhand, had a far-UV spectrum that was significantly diminished in comparison to that of Ca2+-boundaLA. The difference between the two exceeded the expected contribution (theoretically and experimentally determined) of two broken disulfide bonds (Kuwajima et al., 1990, and references cited therein), indicative of a diminution in secondary structure content. There was also a significant change in the near-UV spectrum that could not be attributed solely to the loss of two disulfide bonds. The change was consistent with the removal of some of the chromophores from their asymmetric environments, and the spectrum still contained some fine structure (Figure 6), indicative of the contribution of more than one constrained chromophore [see for comparison, Craig et al. (1989)]. As all the 2SScam molecules had bound Ca2+under these conditions, the CD spectra indicate that only part of the molecule was folded. Removal of Ca2+ from aLA, by the addition of EDTA, resulted in approximately uniform l / 5 reduction in the nearUV CD spectrum, reflecting a decrease in the amount of tertiary structure. In contrast, the ellipticity at 222 nm, indicative of helical content, was hardly changed (Figure 6). These observations are consistent with the results of others [e.g., Kuwajima et al. (1990) and Hiraoka and Sugai (19891 and indicate that apo-aLA was in an equilibrium with 80% of the molecules in the native conformation and 20% in the MG.
The apo form of 3SScam had a near-UV CD spectrum that was much reduced compared to the Ca2+-boundform, but its far-UV dichroism was substantially retained (Figure 6). Similar changes were observed upon removal of ligand from both bovine and human 3SScm (Kuwajima et al., 1990;Ewbank & Creighton, 1991; Ikeguchi et al., 1992). These results are consistent with ap0-3SScam being largely in a MG-like conformation. Removal of Ca2+ from 2SScam resulted in a far greater diminution of CD signal in the near-UV than in the far-UV, but the change in the latter was also significant (Figure 6 ) . Thus, as judged by CD, ap0-2SSCam possessed no tertiary structure and less secondary structure than the Ca2+-bound protein, consistent with partial population of the MG conformation. The rearranged two- and three-disulfide forms all showed essentially featureless near-UV CD spectra (Ewbank, 1992). Their spectra in the far-UV were similar in shape to those of apo 3SScam and 2SScam, and all were more negative than that of Ream (Figure 7). The spectra were not significantly altered by 10 mM Ca2+. The difference in the degree of unfolding measured by changes in the near- and far-UV CD spectra for these derivativesof a-lactalbumin is indicative of the existence of partly folded conformations intermediatebetween the native and unfolded states. A fuller interpretation of these spectra is given below. The conformation of the fully reduced protein and the influence of blocking groups was investigated by comparing the CD spectra of R, R6/l2Ocam, R;(';YY:rm, Itcam, and Rcm (Figures 6 and 7). The minimal far-Uk and featureless nearUV CD spectrum of Rcm(Figure 6) was typical of an unfolded protein and was identical to that previously reported for R c m in 8 M urea (Ikeguchi & Sugai, 1989), indicating that Rcm was fully unfolded even in the absence of denaturant. By contrast, as judged by their similar ellipticities in the far-UV,
Biochemistry, Vol. 32, No. 14, 1993 3701
a-Lactalbumin's Disulfide Folding Intermediates 2.000L...
,
.
.
.
I
I
.
.
I
.
,
.
I
.
.
.
I
.
, ' ]
0
Table 11: Effect of CaZ+on the Intrinsic Fluorescence Parameters of aLA and Its Trapped Disulfide Derivatives" (nm) relative emission intensity species 10 mM Ca2+ 1 mM EDTA 10 mM Ca2+ 1 mM EDTA 323 332 32 42 3 24 339 34 59 335.5 338.5 59 68 340 340 16 80 340 340.5 93 93 340.5 340.5 84 86 341.5 113 341.5 113 96 338.5 338.5 96 345.5 345.5 99 100 The buffer was 0.1 M Tris (pH 8.7), 0.2 M KCI at 25 "C, and the excitation wavelength was 280 nm. Amax
-2.000 -4.000
-6,000
'3
-8.000
d
-10,000
"d
6
Y
0
-2.000 -4.000 -6,000
-8,000 -10,000 -12.000
1'
'
200
.
' *
"
"
240
220
.
" '
'
" '
260
280
'
'
"
300
.' 320
Wavelength (mu)
FIGURE6: Far- and near-UV C D spectra of a L A and its trapped derivatives in the presence and absence of Ca2+ (upper and lower panel, respectively). The samples were a L A (thick line, 6), 3SS,,, 5), 2SScam (- -, 4), R,,, (- - -,3), R,, (- -,l), and a L A in 8 M urea (thin line, 2). The near-UV spectra of R,,, and R,, werevirtually identical to that of a L A in 8 M urea and have been omitted for the sake of clarity. In each case, the buffer was 0.1 M Tris (pH 8.7), 0.2 M KC1, and either 10 mM CaC12 or 1 mM EDTA, a t 25 OC.
-
(.e-,
0
-2,000 -4,000 -6,000 -8,000
I
z"
-10,000 -12,000
-
200
220
240
Wavelength (nm)
FIGURE7: Far-UV CD spectra of rearranged derivatives of a-lactalbumin and fully reduced a-lactalbumin. The rearranged derivatives were in 0.1 M Tris (pH 8.7), 0.2 M KCl, 1 mM EDTA a t 25 OC. Symbols: open triangle, [3SS],a,; open circle, [2SS]cam; closed triangle, [3SS]ia,; closed circle, [2SS]g,. The spectra of reduced a-lactalbumin were recorded under the same conditions, but in the presence of 10 mM DTTSH.thin line, 1 , R in 8 M urea; thick The spectrum of R6/120cam line, 2, R; thin line, 3, R $ k , . overlapped with that of R and has been omitted for the sakeofclarity. The spectra of Ca2+-bound a L A (thin line, 4) and R,, (-) are included for comparison.
R, R6/120cam, R6/IZOcam 28 I l l c a m , and K,,, all possessed some secondary structure (Figures 6 and 7), but their near-UV spectra were virtually zero, indicativeof an absence of tertiary structure (Ewbank, 1992). This negative ellipticity in the far-UV was
virtually abolished by the addition of denaturant; in 8 M urea, their spectra were indistinguishable from that of k,,as illustrated for R in Figure 7. Thus, reduced a-lactalbumin with all eight free thiol groups, or with two, four, or all eight modified by reaction with iodoacetamide, adopted similar partially-populated MG conformations. While blocking the thiol groups with neutral cam groups did little to alter the conformation, complete carboxymethylation caused the protein to be fully unfolded. The spectra of none of the reduced forms, whether blocked or not, were altered by the presence of 10 mM Ca2+(Figure 6; Ewbank, 1992), consistent with none of the reduced forms of a-lactalbumin binding Ca2+. Intrinsic Fluorescence. The intrinsic fluorescence of aLA reflects primarily the environments of its four tryptophan residues (residue numbers 26, 60, 104, and 118). The and relative fluorescence emission were parameters A, determined for aLA and its derivatives, in the presence and absence of Ca2+(Table 11). &, appeared the most unfolded species; its, , ,A was at the longest wavelength, approaching that of free tryptophan (348 nm), while the spectrum of native Ca2+-bound aLA was the most blue-shifted. The native conformation also quenched the fluorescence emission by a factor of 3, comparing Ca2+-boundaLA to &,. The other spectra can be compared with these two extremes. Ca2+bound 3SSc,, appeared on this basis to be almost fully folded. The spectrum of Ca2+-bound2SS,, was intermediate between the two extremes, in terms of both, , ,A and emission intensity. The spectra of the other forms were red-shifted relative to Ca2+-boundaLA, but only by around 3 / 4 of the shift seen for Rem, suggesting that none was as unfolded as &,. Similar trends were apparent in the fluorescence intensity but, curiously, the relative fluorescence intensity of [2SS]:a, exceeded that of Rem. ANS Binding. The hydrophobic fluorescent dye ANS has been used as a diagnostic for the existence of molten globule states in equilibrium and kinetic studies (Ptitsyn et al., 1990; Semisotnov et al., 1991). Fully unfolded and fully folded states show relatively low affinity for ANS compared to partially-folded, molten globule states. The enhancement of ANS fluorescence by denaturant-induced molten globule states, relative to the low level of fluorescence seen with native or fully unfolded states, has been reported to be between 2and 10-fold (Semisotnov et al., 1991). The results of a titration of 50 pM ANS with aLA and its derivatives in the absence and presence of Ca2+are shown in Figure 8. Linearity of fluorescence intensity as a function of protein concentration is a good indicationthat the dye is binding to the monomeric protein and that the protein is not aggregated (Mulqueen & Kronman, 1982). A linear correlation between protein concentration and ANS fluorescence was obtained in all cases, except for the rearranged derivatives. They all had
3102 Biochemistry, Vol. 32, No. 14, 1993
Ewbank and Creighton 3 S b 2sscam
aLA
RRcm
aLA
3 S b 2 S b
R-
Rcm
(PM)
FIGURE8: Binding of A N S by LYLAand its blocked derivatives in 0.1 M Tris (pH 8.7), 0.2 M KCl at 25 "C. (a) Titration of 50 pM A N S with a L A and its blocked derivatives in the presence of 1 mM EDTA. The dashed line shows the nonlinear curve of [3SS]" . The other rearranged derivatives had a relative fluorescence of between 12 and 15 a t 25 p M , but values less than those of 3SSc,, and 2?%,, at lower protein concentrations, and so have been omitted for the sake of clarity. (b) Effect of Ca2+on ANS fluorescence of LYLAand its blocked derivatives. Bars: open, in the presence of 1 mM EDTA; hatched, 10 mM CaC12. Table 111: Contribution of a-Lactalbumin's Native Disulfide. Bonds to the Stability of Its N, MG, and U Stateso conformations present electrophoretic folding equilibria in apo form species Ca2+-boundb amc Kdca (MI mobility of MG stated AGN-u~(kJmol-I) AGMG-~I (kJ-mol-I) aLA 100%N 80% N, 20%MG 2 X 10-8f -9.78 -6.7g 3SSc,, ~100% N 80%MG, 20%U 2.5 X 10-6 97.5 1.7 -3.4 2SSc,, ~100% NPh 60%MG, 40%U 2 X lt5 97 6.9h -1 .o >21.1 30%MG, 70%U > 98 2.1 Rcm na 0.1 M Tris (pH 8.7), 0.2 M KCI at 25 O C . In the presence of 10 mM ea2+. In the presence of 1 mM EDTA; values considered accurate to &lo%. Calculated from the observed electrophoretic mobility of each apo form (Table I) by Mobil,, =fMGMObilMG +fuMobilu; wherefu andfMG are the fractional populations of U and MG states, respectively, and Mobilu is given by the observed mobility of the corresponding form in 8 M urea (TableI). e Calculatedon theassumption that theobservedaffinityfor ea2+reflectsonly thepopulationoftheNstate;AGN-u =-9.7-RTln(Ka/5XlO7 M-I). f Estimated from values of Hiraoka and Sugai (1985) for the conditions used here. Xie et al. (1991); 10 mM borate (pH 8.0), 1 mM EDTA, 0.2 M NaCI, 25 O C . These values give AGN-MG= -3.0 kJ.mol-l, which compares well with the value of -3.4 kJ.mo1-' obtained from the population of N and MG states of apo-aLA here. N,, native in part, remainder in collapsed unstructured state; see text for details. i Value for unblocked 2SS (Ewbank & Creighton, 1993).
substantial ANS fluorescence, which was increased in the aggregated state at protein concentrations greater than 10 rM. Ca2+-boundaLA and R,,, which by the structural criteria detailed above were fully folded and fully unfolded, respectively, had the expected low affinity for ANS. The greater fluorescenceof all the other forms of a-lactalbumin suggested that all had some molten globule character. The dramatic difference between the ANS fluorescence observed for apoand Ca2+-bound 3SS,,, enabled its dependence upon Ca2+ concentration to be measured (Ewbank, 1992). The apparent dissociation constant for Ca2+ (2.5 X 10-6 M) was 10-fold greater than that estimated from the effect of changing the Ca2+ concentration on the rate of reduction of 3SS measured in the preceding paper (Ewbank & Creighton, 1993).
DISCUSSION aLA has been extensively studied recently in efforts to understand why it can exist under certain conditionsin a stable partially-unfolded MG state, in addition to the fully folded, N, and fully unfolded, U, states (Kuwajima, 1989; Alonso et al., 1991; Ewbank & Creighton, 1991, 1993;Xieet al., 1991; Ikeguchi et al., 1992). Following the finding that a threedisulfide derivativeof human a-lactalbumin could exist in the MG state at neutral pH (Ewbank & Creighton, 1991), the pathways and kinetics of disulfide bond reduction and reformation of bovine aLA were examined in far greater detail under conditions favoring the different conformational states, and various kinetic intermediates were identified (Ewbank & Creighton, 1993). To aid interpretation of thesekinetic studies and to further understanding of a-lactalbumin's different conformational states, structural studies have been carried out on the trapped reduction intermediates under the same
conditions as used for the disulfide folding studies. A variety of techniques that give complementary information have been used. Additional information for each species is given by its affinity for Ca2+,by the disulfides it contains, by the rates at which these disulfides are rearranged and broken, and by the rates at which further disulfides are formed. In this way, a consistentdescription has been generated of the conformational properties of a-lactalbumin in terms of the N, MG, and U conformational states (Table 111). Disulfide Bonds of 3SS and 2SS. The native disulfide bond Cys6-120 has been confirmed to be that reduced rapidly in aLA and absent from 3SS (Shechteret al., 1973;Kuwajima et al., 1990). The free thiols in 2SS have been identified as those of Cys6, Cysl20, Cys28, and Cyslll, consistent with the Cys28-111 disulfide being the second to be reduced, directly in 3SS. The identity of the disulfides remaining in 3SS and 2SS could not be established directly because of the difficulty of obtaining proteolytic cleavage between the near neighbors Cys73 and Cys77. Indirect evidence, however, strongly supports the retention of the other native disulfide bonds. Re-formation of the Cys6-120 disulfide in 3SS regenerated native aLA. The reduction of Ca2+-bound3SS to 2SS occurred through the folded conformation and must have involved cleavage of one of the native disulfide bonds. Although it is theoretically posible that the partners in the two remaining disulfide bonds could have been swapped by a series of rapid thioldisulfide rearrangement reactions (a minimum of three steps), Ca2+-bound2SS is relatively stable and there were no indicationsof disulfidebond rearrangements having occurred (Ewbank & Creighton, 1993). Additionally, the two remaining disulfide bonds of 2SS flank aLA's Ca2+ binding site, which would undoubtedly be disrupted by their rearrangement (see Figure 1 in the preceding paper; Ewbank
a-Lactalbumin's Disulfide Folding Intermediates & Creighton, 1993). Both 3SS and 2SS bind Ca2+ with relatively high affinity and, in the absence of any evidence for disulfide bond rearrangements, it is very likely that 3SS differs from aLA only by the absence of the Cys6-120 disulfide, that 2SS lacks only this disulfide and Cys28-111, and that both retain the remaining native disulfide bonds. Effect of the Blocking Groups. The kinetic intermediates had to be trapped covalently in order to study them under conditions like those used in the studies of disulfide bond reduction, rearrangement, and re-formation. Iodoacetamide was found to be efficient and specific in blocking the free thiols of 3SS, 2SS, and R. For [3SS] and [2SS], however, many different disulfide isomers are in rapid equilibrium, so there is the possibility that irreversible blocking might perturb the equilibrium mixture if the thiol groups of different species have different reactivities, but similar HPLC profiles were obtained of rearranged species trapped with various high concentrations of iodoacetamide (results not shown). The introduced blocking groups might affect the conformation and stability of the protein due to their steric bulk and polar groups, as discussed recently by Ikeguchi et al. (1992). Ionization of cm or thiol groups would be expected by electrostatic repulsion to inhibit collapse of an acidic protein like aLA, which may be the reason that is apparently fully unfolded. At low pH, the cm groups will be neutralized, which may explain the observation of secondary structure by far-UV CD in Rc, at pH 2 (Ikeguchi & Sugai, 1989). The opposing factors of partial ionization of free thiols groups and steric bulk of cam groups appeared to balance or to be minimal with the reduced protein: R, R6/i20cam, R6,$~~8,~,,and R,, have 0, 2, 4, and 8 thiols blocked with cam groups, and the others as free thiols. All had similar conformations, with -70% of the molecules in the U state and the remainder in the MG state, as judged by their CD spectra. The effects of the blocking groups would be expected to be different in the fully folded state. The structure of trapped Ca2+-bound 3SS was the same whether iodoacetamide or iodoacetic acid was used to block its two free thiol groups (Figure 4; Ikeguchi et al., 1992). For both 2SS and 3SS, urea gradient electrophoresis unfolding transitions of the cam and cm derivatives occurred over a similar denaturant range (results not shown). The rate of reduction of Ca2+-bound 2SScam by DTTZ; was, however, somewhat slower than that of the same protein with four free thiol groups [0.2 s-l M-l as against 1.4 s-l M-'1, and a slight effect was observed with 3SS (Ewbank & Creighton, 1993). This presumably reflects a small net stabilizing effect of the cam group relative to the partially-ionized thiol. Apart from this small effect on stability, the cam derivatives used here appear to be close structurally to the unblocked reduction intermediates. Aggregation of Partially Folded Proteins. Partially- folded proteins, especially molten globules that expose an increased hydrophobic surface relative to their folded forms, are prone to aggregation, which affects their physical properties and makes difficult their characterization. Aggregation is known to occur with a-lactalbumin (Kronman et al., 1964; Ikeguchi & Sugai, 1989), but its occurrence is often not apparent with many techniques. Aggregation in the studies described here usually involved formation of intermolecular disulfide bonds and was then very obvious by gel electrophoresis, as even dimeric complexes have greatly reduced electrophoretic mobilities (Ewbank & Creighton, 1993). At protein concentrations greater than 100 pM, there was also spectroscopic indicationsof time-dependentaggregation of all the derivatives, with the exception of the fully folded Caz+-bound aLA and 3SScam. Even at submicromolar concentrations, ANS pro-
Biochemistry, Vol. 32, No. 14, 1993 3703
motes the aggregation of aLA at acid pH (Mulqueen & Kronman, 1982; Semisotnov et al., 1991), and here at 50 pM promoted the aggregation of the rearranged derivatives (Figure 8); in the absence of ANS, however, they behaved as monomeric species at low concentrations. Aggregation was found not to be a significant problem in these studies when the protein concentration was kept below 25 pM and is very unlikely to have affected the results presented here. Ca2+Binding. A single tightly-bound Ca2+ion specifically stabilizes the native conformation of a-lactalbumin (Acharya et al., 1989, 1991). The native conformation is far less temperature stable in apo-aLA than when CaZ+is bound (Kuwajima et al., 1986), and the disulfides are much less stable and are reduced rapidly (Ewbank & Creighton, 1993). This effect of CaZ+makes it possible to stabilize the native conformation specifically by varying the Ca2+concentration. Fully reduced a-lactalbumin and the derivatives with some of the native disulfide bonds have the potential, however energetically unfavorable, to adopt an N-state conformation that would differ from aLA's only in the number of native disulfide bonds. None of the disulfide bonds are directly involved in the Caz+-bindingsite, so each N-stateconformation would be expected to bind Ca2+with the same affinity, KdCa, that of apo-aLA where the N conformation predominates. So long as only specific binding by the N conformation is measured, the stabilizing effect is directly proportional to the free Ca2+ concentration when the protein is fully saturated. This can be illustrated with the following equilibria for any species i able to adopt the N conformation: Kr.i
Ui-Ni
-
KdCa
Ca2+
N*Ca2+
(1)
where Kf,iis the equilibrium constant for folding of the apo form of species i and Kdca is the dissociation constant of the native conformation for CaZ+, which is assumed to be independentof the species. The apparent equilibrium constant for folding of species i in the presence of Ca2+,Kf,app,,,will be given by
When the protein is saturated, at high CaZ+concentrations, the stability is directly proportional to the free CaZ+concentration: (3) The N state of a-lactalbumin is stabilized not only by Ca2+, however, but also by a variety of monovalent and divalent metal ions and by Tris (Permyakov et al., 1985; Hiraoka & Sugai, 1985), so quantitative comparisons of the Caz+-free form under varying conditions in different studies are uncertain. The data of Hiraoka and Sugai (1985) indicate that the value of KdCafor aLA is approximately 2 X lo-* M under the conditions used here (Ewbank, 1992). In this case, the N state should be stabilized by 32 kJ/mol by the presence of 10 mM free CaZ+ and should become populated to a significant amount in any species if it is present in the apoprotein to an extent greater than 106. When the nativeconformationof any speciesis not populated sufficiently in the absence of CaZ+to be detectable directly, its stability can be estimated from the observed affinity of the speciesfor CaZ+.The apparent dissociation constant for Ca2+
3704 Biochemistry, Vol. 32, No. 14, 1993
Ewbank and Creighton
of a species i obeying eq 1 will be given by
Where the N state is not substantially populated in the apo form, this simplifies to Kdc"
Kf,) =-
(5)
Kd,app,i
With KdCa = 2 X lo-* M and concentrations of Ca2+of up to 0.01 M, the presence of the N state can be quantified over 6 orders of magnitude. If a species i does not detectably bind Ca2+at a free concentration of 10mM, the native conformation is populated in the apo form of i to an extent no greater than 10". The above discussion treats the N state as a cooperative structure, as is usually the case, but a-lactalbumin is somewhat unusual in that only part of the native conformation can be present, as in Ca2+-bound2SS (see below). Indeed, the Ca2+ binding site of aLA consists of only residues 79-88, in a short segment of polypeptide chain, so the above discussion can be modified to include the possibility of just a local Ca2+binding site existing. Structure of Apo-aLA. The results presented here suggest that dissociation of Ca2+ from aLA causes the MG state to be populated in a fraction of the molecules, approximately 20% as judged by the near-UV CD, rather than causing a local perturbation of the N conformation. The presence of the MG state would account for the change in the physical properties of aLA (Permyakov et al., 1985; Hiraoka & Sugai, 1985) and would be reflected in the 5-fold enhancement of ANS binding and the alteration of intrinsic fluorescence properties. The change in intrinsic fluorescence has been attributed to a change in the environment of all emitting tryptophans (Ostrovskyet al., 1988), with a majorcontribution from Trpl18 (Murakami et al., 1982), which is distant from the Caz+-binding loop. This conclusion is also consistent with the greatly increased susceptibility of the apoprotein to proteolytic digestion (Hirai et al., 1992) and with a recent calorimetricstudyofapo-bovineaLA(Xieetal., 1991), which concluded that the MG state is some 15% populated in apoaLA in the absence of denaturant under conditions reasonably similar to those used here. Structure of 3SSCam.With Ca2+ bound, 3SScam adopts a near-fully folded conformation: its electrophoretic mobility was indistinguishable from that of aLA, it had native-like CD and fluorescence spectra, and its remaining disulfide bonds were reduced relatively slowly (Ewbank & Creighton, 1993). Ikeguchi et al. (1992) have reached the same conclusion. The relative ANS fluorescence of Ca2+-bound 3SSCam was, however, found here to be comparable to that of apo-aLA, which as described above appears to exist in a partially populated MG state. This ANS binding may be the consequence of a local relaxation in the region of the cleaved disulfide bond. The thiol groups of Cys6 and Cysl20 in Ca2+-bound 3SS tended to interchange slowly with the Cys28-111 disulfide, reflecting a substantial dynamic flexibility in the molecule. For the analogous three-disulfidederivativeof lysozyme, minor conformational rearrangements occurred in the C-terminal region (Radford et al., 1991). The absence of a disulfide bond may also influence the quantum yield of the bound dye or the number of available ANS binding sites. Ap0-3SScam had all the characteristics of a molten globule (Kuwajima, 1989) under the conditions used here. It was comparatively compact and possessed considerable secondary
structure as judged by CD. It displayed a dramatic enhancement of ANS fluorescence, had an intrinsic fluorescence spectrum distinct from that of aLA in either the N dr U state, and did not have significant tertiary interactions as judged by its near-UV CD spectrum and the rapid rate at which the disulfide bonds of the unblocked form rearranged (Ewbank & Creighton, 1993). The N state of apo-3SS predominates at low temperature, however; reduction of apo-aLA essentially stopped at the 3SS stage at 1 OC (Iyer & Klee, 1973), and apo-3SScamat 4 OC had a CD spectrum like that of the native protein (results not shown). Partly-Folded Structure of 2SSCam. The results presented here indicate that part of Ca2+-bound2SSMm was in a folded conformation under the standard conditions: it retained aLA's Ca2+binding site, it had a far-UV CD spectrum indicative of some helical structure, its two remaining disulfide bonds were somewhat resistant to reduction (Ewbank & Creighton, 1993), it underwent a moderately cooperative urea-induced unfolding transition, and it was as compact as aLA (Table I). It exhibited significant ANS binding, however, and had substantially altered intrinsic fluorescence and CD spectra relative to folded aLA. The corresponding unblocked form, Ca2+-bound2SS, preferentially formed nonnative disulfide bonds, while the rate of formation of the Cys28-111 disulfide was of a similar order of magnitude to that of a disulfide in a fully unfolded protein (Ewbank & Creighton, 1993). Such observations could be consistent with a three-state equilibrium between molecules in which some of the molecules are fully N, fully MG, or fully U, as has been assumed with the other species described here, but this is unlikely in the case of Ca2+-bound2SS: all of the molecules were saturated with Ca2+, so an equilibrium was unlikely. An explanation for these findings is that the region of the molecule around the Ca2+binding site, encompassing at least residues 61-91 (see Figure 1 in the preceding paper; Ewbank and Creighton, 1993), exists in a folded conformation when Ca2+ is bound, while the remainder is in a collapsed but unstructured conformation. That the disordered part of 2SS is collapsed, but does not appear to have secondary structure, suggests that the two phenomena may not need to occur concomitantly. The absence of secondary structure in the unfolded part of 2SS also suggests that there may need to be a critical amount of disordered polypeptide chain for the collapsed MG state with secondary structure to be stable. This partly-folded structure of Ca2+-bound2SSMmcontrasts with the usual cooperativity of protein unfolding transitions and demonstrates that part of the protein can unfold independently. There is supporting evidence from X-ray crystallography for local, partial unfolding in native human a-lactalbumin (Harata & Muraki, 1992). This change, a helix-loop transition that was suggested to be caused by the protonation of Hisl07, occurred just in the region that would be predicted to be affected by reduction of the Cys28-111 disulfide. Thus, aLA appears to consist of two structural subdomains, which correspond only partly to those defined crystallographically (domain 1, residues 1-37 and residues 86-123; domain 2, residues 38-85; Acharya et al., 1991). It should be noted, however, that the two halves are not totally independent, for reducing the Cys28-111 disulfide bond to produce 2SS increased greatly the rate of reduction of the remaining two disulfides in the other part of the molecule (Ewbank & Creighton, 1993). The closest counterpart of Ca2+-bound 2SS may be the one-disulfide folding intermediate of BPTI containing only the Cys30-51 disulfide bond (van Mierlo et al., 1992, 1993; see below). The structure of Ca2+-bound2SS is reminiscent
a-Lactalbumin’s Disulfide Folding Intermediates of the critical substructure model of Kuwajima et al. (1989), which was proposed as a high-energy transition state between the MG and N states. In contrast, Ca2+-bound2SSca, has a stable equilibrium structure, although trapped as a kinetic intermediate. The ability to generate stable partly-folded structures when only some of the native disulfides are present demonstrates the power of the disulfide approach to studying protein structure and folding. The structure of Ca2+-bound2SScam may also be compared with the suggested transient folding intermediate of the homologous lysozyme, which was proposed to resemble the equilibrium MG state of aLA and which is split into two subdomains, one structured and the other not (Miranker et al., 1991;Radfordet al., 1992). The presumed transient folded subdomain in lysozyme overlaps with, but is distinct from, the proposed native-like part of Ca2+-bound2SScam. Thus, the major C-helix, which includes Cys91 that participates in a disulfide bond resistant to reduction in Ca2+-bound2SSca,, is also present in lysozyme’s early folding intermediate. In contrast, the preceding 3]0 helix (residues 76-82 in aLA), which is presumed also to be structured in Caz+-bound2SSca,, on the basis of the resistance of the Cys61-77 disulfide bond to reduction, lies in tHbunstructured subdomain of lysozyme’s early folding intermediate (Miranker et al., 1991; Radford et al., 1992). This differencemay be a consequence of the absence of a Ca2+-bindingsite in lysozyme, sincethe bound Ca2+greatly stabilized the folded conformation of 2SS and it was stable only at high Ca2+concentrations. Removal of Ca2+ from 2SSca, resulted in comparatively small changes in the protein’s intrinsic fluorescence and ANS binding properties. There was, however, a Substantial reduction in the protein’s compactness and stability. Its remaining disulfide bonds were reduced 40 times more rapidly than in the Ca2+-boundprotein (Ewbank & Creighton, 1993), and its near-UV CD spectrum was absent. Hence no part of apo-2SScamappeared to be in a folded conformation. Apo2SScam did have a marked far-UV spectrum, however, and underwent a gradual unfolding transition. The overall helical content of 2SSca,, as judged by the intensity of its CD signal at 220 nm changed little upon removal of ligand, but there was a Substantial change in the shape of the spectrum, similar to that seen for 3SSca,. These observed properties of apo2SScam are consistent with it being in an equilibrium between approximately 60% of molecules in the MG state and 40% in U. Cleavage of the Cys28-111 disulfide therefore seems to shift the conformational equilibrium from the MG toward the U state. This would also be consistent with the greater overlap, relative to 3SS, between the products of disulfide bond rearrangement of apo-2SS in the absence and presence of urea, [2SS] and [2SSIu, respectively (Figure 4). Comparison with the BPTI Intermediates. Parallels can be drawn between the structures of a-lactalbumin’s trapped intermediates and the disulfide bond intermediates that populate the folding pathway of BPTI [reviewed by Creighton (1978), Goldenberg (1992), and van Mierlo et al. (1992)l. Thus, intermediates (30-51,5-55), (30-51,14-38), and (555, 14-38), which each lack one native disulfide bond, and the one-disulfide species (5-55) all adopt conformations of varying stability that are extremely native-like, probably due to the high stability of the N state of BPTI; they resemble Ca2+-bound3SS of a-lactalbumin. Intermediates in which the only native disulfide is Cys30-5 1, i.e., (30-5 l), (30-5 1, 5-14),and (30-51,5-38), all havesimilar structures in which the main elements of native secondary structure are intact, but the N-terminal fourth of the molecule is disordered (Darby et al., 1992). Consequently, these species are analogous to
Biochemistry, Vol. 32, No. 14, 1993 3705
Ca2+-bound2SS. In the case of a-lactalbumin, a Substantial part of the stabilization of the folded part of the molecule comes from Ca2+binding, whereas in the BPTI derivatives it is provided primarily by nonpolar interactions between elements of secondary structure and by a disulfide bond. Both cases, however, illustrate that thecontribution ofan interaction can be local, even in a protein whose structure exhibits cooperativity when fully folded (Creighton, 1978; Goldenberg et al., 1992). In contrast to the disordered part of CaZfbound 2$S, the disordered part of the analogous BPTI intermediatesis not collapsed (van Mierlo et al., 1992,1993). Neither BPTI nor its derivatives lacking native disulfide bonds appear to adopt a stable MG state, unlike a-lactalbumin. The precise physical basis for such a distinction is an outstanding question in the study of MG states. Structures of the RearrangedSpecies. When disulfidebond partners may exchange, the most prevalent pattern of disulfide pairing will reflect the most stable conformation, as was first demonstrated for hsulin and ribonuclease A (Givol et al., 1965). If 3SS or 2SS are fully unfolded by the addition of 8 M urea, their remaining disulfide bonds are not stable, but interchange rapidly with the free cysteine thiol groups to give many rearranged forms, with no detectable nonrearranged species remaining. Even in the absence of any denaturant, the folded conformations of 2SS and 3SS are destabilized upon removal of CaZ+,they adopt the MG and U conformations, and their disulfide bonds rearrange, just as rapidly as when they are fully unfolded in 8 M urea. Less than 4% and 2% of 3SS and 2SS, respectively, would be expected to remain at equilibrium, on the basis of the ratios of the forward and reverse rate constants for their rearrangement (Table I1 in the preceding paper; Ewbank & Creighton, 1993), and none were detected. As they were generated only upon removal of Ca2+,the rearranged species cannot bind Ca2+as tightly as the species 3SS and 2SS, and there was no effect of Ca2+ on the rearranged products’ conformational properties. Disulfide rearrangements of apo-3SS and apo-2SS in the presence and absence of urea produced different mixtures of intermediates. The average properties of the mixtures [3SS]cam, [3SS]:a,, [2SS]cam, and [2SS]:a, in the absence of urea were distinct from those expected for a mixture of completely unfolded proteins; all have some secondary structure as judged by CD, have an intrinsic fluorescence spectrum blue-shifted relative to the fully unfolded Rem, have a Substantial affinity for ANS (which promotes their aggregation), are partially collapsed, and undergo denaturantinduced unfolding transitions of low cooperativity. They thus possess some MG-like character in the absence of denaturant which, on average, appears to decrease in the order [3SS],,, > [2SS]cam = [3SS]:a, > [2SS]:a, as the amount of unfolded character increases. The previous conclusion that the iodoacetate-trapped three-disulfide species obtained after rearrangement of human 3SS in the presence of 8 M urea remained unfolded when the denaturant was removed (Ewbank & Creighton, 1991) has been reassessed in light of the current findings on bovine [3SS]:a,. The former species exhibit marked ANS binding (Ewbank, 1992) and are concluded also to have some MG-like character in the absence of urea. The rearranged species exhibited a range of electrophoretic mobilities (Figures 4 and 5 ) . In the absence of denaturant, a small proportion of [3SS]ca, surprisingly had a mobility greatbr than that of 3SSam and comparable to that of apoaLA. Unless disulfide rearrangement causes very Substantial alterations in the pKa values of ionized groups, some of the 3SS molecules must upon rearrangement either become even more compact or undergo a radical change in overall shape.
3706 Biochemistry, Vol. 32, No. 14, 1993 Stabilization of the N State by Disulfide Bonds. By quantifying the extents to which the N and U states are populated in the apo forms of thevarious species with different disulfide bonds, the contribution of thedisulfides to the stability of the N state can be estimated (Table 111). The Cys6-120 disulfide stabilizes aLA's state by 11.4 kJ/mol, as calculated from the difference in the free energy between their respective U and N states for aLA and 3SScam; this is in satisfactory agreement with the difference of 10.1 kJ/mol between aLA and 3SS,, measured by Kuwajima et al. (1990). This stabilization is presumably decreased by about 10 kJ/mol by the apparent strain in the Cys6-120 disulfide bond (Kuwajima et al., 1990; Ewbank & Creighton, 1993). The energetic contribution of the Cys28-111 disulfide bond to the stability of the N state is considerably less (4 kJ/mol), but this is a minimum value for the entire native conformation, for only part of the CaZ+-bound 2SS appeared to be folded. Values of 1 7 kJ/mol are obtained for the other two native disulfides, assuming that they are equivalent. Reducing the disulfide bonds of a-lactalbumin therefore destabilizes substantially the N state, so that in the absence of CaZ+ it is no longer populated substantially after more than one disulfide bond is cleaved and occurs to < 1 V in the fully reduced protein, R. The lowered population of the N state causes the apparent affinity of the protein for Ca2+ to be decreased accordingly (eq 5 ) . Molten Globule States. The observed properties of aLA and its disulfide derivatives were consistent with a similar MG state for each, at least in terms of secondary structure content and compactness (Table 111). The relative electrophoretic mobilities of the MG states of 3SScam, 2SScam, and Rca, could be calculated to estimate their hydrodynamic volumes (Table 111) and were found to be very similar, with an average value of 97.5. This is only slightly less than the mobility of the N state (100.6) and considerably greater than that of aLA in the U state with the four native disulfide bonds intact (78) and without any disulfides (48). These values are consistent with the similar compactness of the N and MG states of aLA reported previously (Kuwajima et al., 1989). Consequently, the unfolding transition monitored by ureagradient electrophoresis, in the absence of Ca2+,is primarily that between the MG and U states. The urea-induced unfolding transitions of apo-aLA, apo-3SSa,, and apo-2SSam and of R,,, werevery gradual (Figure 2), indicative of limited structural cooperativity of their MG states. The native disulfide bonds contribute much less to the stability of the MG state than they do to that of the N state (Table 111). The contribution of each was -2-3 kJ/mol; consequently, the MG state was still populated in R, in the absence of any disulfide bonds. A similar conclusion was reached by Ikeguchi et al. (1992) for the Cys6-120 disulfide bond. For the MG state, the stabilization is nonspecific: when either 3SS or 2SS is in a MG conformation, their native disulfides are not stable; they undergo rapid thiol-disulfide interchange to give a large number of rearranged species with more stable disulfide bonds, which can contain little or no common specific structure, but appear to maintain a MG-like conformation. The products of the rearrangement of apo3SS appear less unfolded than the corresponding two-disulfide forms. Thus, the small general contribution of disulfide bonds to the increased occupancy of the MG state may result simply from the additional conformational constraint contributing to the maintenance of a compact molecule (Dill & Shortle, 1991). The conformational entropy of the MG state is presumably reduced by disulfide bonds only slightly less than is that of the U state.
Ewbank and Creighton Certain elements of native secondary structure, for example part of the major C-helix, are retained at acid pH in the MG state of disulfide bond-intact guinea pig a-lactalbumin, and there is evidence for the persistence of a native-like clustering of aromatic residues (Baum et al., 1989). At least some of these specific features of aLA's MG state may be the result of the retention of the four native disulfide bonds. The energetically preferred rearrangements of the disulfides in the MG state are extensive and involve all the cysteine residues (Ewbank & Creighton, 1993), which reduces the likelihood that any specific super-secondary contacts are retained in all the rearranged species. With fewer of the fixed constraints provided by native disulfide bonds to maintain the overall native fold, the MG state is likely to be a more dynamically flexible conformational state. The results presented here support the increasingly accepted view that MG states are stabilized largely by nonspecific hydrophobic interactions (Kuwajima, 1989; Ewbank & Creighton, 1991; Mitaku et al., 1991; Hughson & Baldwin, 1991). In this case, the higher order interactions, above the level of secondary structure, that stabilize the N state would not be expected to be important for the stabilization of transient MG intermediates during protein folding; certainly as judged by the rearranged species, collapse and formation of secondary structure appear not to require an overall native fold. The MG state is, however, nearly as compact as the N state and must consequently serve to limit severely the conformational space that may be sampled by a polypeptide chain during folding; indeed, it substantially increases the rate of initial disulfide bond formation in reduced a-lactalbumin (Ewbank & Creighton, 1993). ACKNOWLEDGMENT
Part of this work was conducted at the MRC Laboratory of Molecular Biology, Cambridge, U.K., when J.J.E. received an MRC studentship. We thankT. Houthaeve and R. Kellner for amino acid analysis, L. J. Berliner for communication of results prior to publication, J. Lakey and T. Gibson for critical reading of the manuscript, and V. Canale, N. J. Darby, P. A. Evans, S.E. Radford, and G. Vriend for discussions. REFERENCES
Acharya, K. R., Stuart, D. I., Walker, N. P. C., Lewis, M., & Phillips, D. C. (1989) J. Mol. Biol. 208, 99-127. Acharya, K.R., Ren, J., Stuart, D. I., Phillips, D. C., & Fenna, R. E. (1991) J. Mol. Biol. 222, 571-581. Alonso, D. 0.V., Dill, K. A., & Stigter, D. (1991) Biopolymers 31, 1631-1649. Baldwin, R. L. (1991) Chemtracts Biochem. Mol. Biol. 2, 379389. Baum, J., Dobson, C. M., Evans, P. A., & Hanley, C. (1989) Biochemistry 28, 7-13. Bolotina, I. A. (1987) Mol. Biol. (USSR)(English translation) 21, 1328-1337. Cowburn,D. A., Brew, K., & Gratzer, W. B. (1972) Biochemistry 11, 1228-1234. Craig, S . , Pain, R. H., Schmeeissner,U., Virden, R., & Wingfield, P. T. (1989) Int. J. Pept. Protein Res. 33, 256-262. Creighton, T. E. (1978) Prog. Biophys. Mol. Biol. 33,231-297. Creighton, T. E.(1979) J. Mol. Biol. 129, 235-264. Darby, N. J., van Mierlo, C. P. M., Scott, G. H. E., Neuhaus, D., & Creighton, T. E. (1992) J. Mol. Biol. 224, 905-911. Davis, B. J. (1964) Ann. N.Y.Acad. Sci. 221, 404-427. Dill, K.A., & Shortle, D. (1991) Annu. Reu. Biochem. 60,795825. Dolgikh, D. A., Gilmanshin, R. I., Braznikov, E. V., Bychkova, V. E., Semisotnov, G. V., Venyaminov, S.Yu.,& Ptitsyn, 0. B. (1981) FEBS Letf. 236, 311-315.
a-Lactalbumin’s Disulfide Folding Intermediates Ewbank, J. J. (1992) Ph.D. Thesis, Cambridge University, England. Ewbank, J. J., & Creighton, T. E. (1991) Nature 350,518-520. Ewbank, J. J., & Creighton,T. E. (1993) Biochemistry (preceding paper in this issue). Gil’manshin, R. I., Dolgikh, D. A,, Ptitsyn, 0. B., Finkel’shtein, A. V., & Shakhnovich, Y. I. (1982) Biophysics 27, 10521064. Givol, D., De Lorenzo, F., Goldberger, R. F., & Anfinsen, C. B., (1965) Proc. Natl. Acad. Sci. U.S.A. 53, 676-684. Goldenberg, D. P. (1992) Trends Biochem. Sci. 17, 257-261. Goldenberg, D. Po,& Creighton, T. E. (1984) Anal. Biochem. 138, 1-18. Goldenberg, D. P., Berger, J. M., Lahern, D. A., Wooden, S.,& Zhang, J.-X. (1992) Biochemistry 29, 5083-5087. Goto, Y., Calciano, L. J., & Fink, A. L. (1990) Proc. Natl. Acad. Sci. U.S.A. 87, 573-577. Harata, K., & Muraki, M. (1991) J. Biol. Chem. 267, 14191421. Hirai, Y., Permyakov, E. A., & Berliner, L. J. (1992) J. Protein Chem. 11, 51-57. Hiraoka, Y . , & Sugai, S. (1985) Znt. J. Pept. Protein Res. 26, 252-26 1. Hiraoka, Y., Segawa, T., Kuwajima, K., Sugai, S.,& Murai, N. (1980) Biochem. Biophys. Res. Commun. 95, 1098-1 104. Hughson,F. M., & Baldwin,R.L. (1991) Biochemistry28,44154422. Ikeguchi, M., & Sugai, S. (1989) Znt. J. Pept. Protein Res. 27, 289-297. Ikeguchi, M., Kuwajima, K., Mitani, M., & Sugai, S. (1986a) Biochemistry 25, 6965-6972. Ikeguchi, M., Kuwajima, K., & Sugai, S. (1986b) J. Biochem. (Tokyo) 99, 1191-1201. Ikeguchi, M., Sugai, S.,Fujino, M., Sugawara,T., & Kuwajima, K. (1992) Biochemistry 31, 12695-12700. Iyer, K. S.,& Klee, W. A. (1973) J. Biol. Chem. 248,707-710. Kronman, M. J., & Bratcher, S. C. (1983) J. Biol. Chem. 258, 5707-5709. Kronman, M. J., Andreotti, R., & Vitols, R. (1964) Biochemistry 3, 1152-1160. Kronman, M. J., Creankowski, L., & Holmes, L. G. (1965) Biochemistry 4, 5 18-525. Kuwajima, K. (1989) Proteins: Struct., Funct., Genet. 6, 87103. Kuwajima, K., Ogawa, Y., & Sugai, S. (1981) J. Biochem. 89, 759-770. Kuwajima, K., Hiraoka, Y., Ikeguchi, M., & Sugai, S. (1985) Biochemistry 24, 874-88 1. Kuwajima, K., Harushima, Y.,& Sugai, S. (1986) Znt. J . Pept. Protein Res. 27, 18-27. Kuwajima, K., Mitani, M., & Sugai, S. (1989) J.Mol. Biol. 206, 547-561. Kuwajima,K., Ikeguchi,M., Sugawara, T., Hiraoka, Y., & Sugai, S.(1990) Biochemistry 29, 8240-8249.
Biochemistry, Vol. 32, No. 14, 1993 3707
Kuwajima, K., Garvey, E. P., Finn, B. E., Matthews, C. R., & Sugai, S. (1991) Biochemistry 30, 7693-7703. Manning, M. C., & Woody, R. W. (1989) Biochemistry 28, 8609-86 13. Miranker, A., Radford, S.E., Karplus, M., & Dobson, C. M. (1991) Nature 349, 633-636. Mitaku, S.,Ishido, S.,Hirano, Y., Itoh, H., Kataoka, R., & Sait6, N. (1991) Biophys. Chem. 40, 217-222. Mulqueen, P. M., & Kronman, M. J. (1982) Arch. Biochem. Biophys. 215, 28-39. Murakami, K., Andree, P. J., & Berliner, L. J. (1982) Biochemistry 21, 5488-5494. Ostrovsky,A. V., Kalinichenko, L., Emelyanenko,V. I., Klimanov, A. V., & Permyakov, E. A. (1988) Biophys. Chem. 30,105112. Permyakov, E. A., Yarmolenko, V. V., Kalinichenko, L. P., Morozova, L. A., & Burstein, E. A. (1981) Biochem. Biophys. Res. Commun. 100, 191-197. Permyakov, E. A., Morozova, L. A., & Burstein, E. A. (1985) Biophys. Chem. 21, 21-31. Ptitsyn, 0.B. (1992) in Protein Folding (Creighton, T. E., Ed.) pp 243-300, W. H. Freeman, New York. Ptitsyn, 0. B., Pain, R. H., Semisotnov, G. V., Zerovnik, E., & Razgulyaev, 0. I. (1990) FEBS Lett. 262, 20-24. Radford, S. E., Woolfson, D. N., Martin, S. R., Lowe, G., & Dobson, C. M. (1991) Biochem. J. 273, 21 1-217. Radford, S.E., Dobson, C. M., & Evans, P. A. (1992) Nature 358, 302-307. Schaer, J. J., Milos, M., & Cox, J. A. (1985) FEBS Lett. 190, 77-80. Semisotnov, G. V., Rodionova,N. A., Razgulyaev,0.I., Uversky, V. N., Gripas’, A. F., & Gil’manshin, R. I. (1991) Biopolymers 31, 119-128. Shechter, Y., Patchornik, A., & Burstein,Y. (1973) Biochemistry 12, 3407-3413. Shewale, J. G., Sinha, S.K., & Brew, K. (1984) J. Biol. Chem. 259, 4947-4956. Stemmer, P., & Klee, C. B. (1990) FEBS Lett. 276, 71-74. Thompson, M. P., Groves, M. L., Brower, D. P., Farrell, H. M., Jenness, R., & Kotts, C. E. (1988) Biochem. Biophys. Res. Commun. 157, 944-948. van Mierlo, C. P. M., Darby, N. J., & Creighton, T. E. (1992) Proc. Natl. Acad. Sci. U.S.A. 89, 6775-6779. van Mierlo, C. P. M., Darby, N. J., Keeler, J., Neuhaus, D., & Creighton, T. E. (1993) J. Mol. Biol. (in press). Vilotte,J.-L., Soulier,S.,Mercier, J.-C., Gaye, P., Hue-Delahaie, D., & Furet, J. P. (1987) Biochimie 69, 609-620. Weber, G., & Young, L. B. (1964) J. Biol. Chem. 239, 14151423. Xie, D., Bhakuni, V., & Freire, E. (199 1) Biochemistry 30,1067310678.